COGNITION AND EMOTION

2012, 26 (7), 1223
1237

Attentional capture by emotional faces is contingent on

attentional control settings
Daniel Barratt and Claus Bundesen
Department of Psychology, University of Copenhagen, Copenhagen, Denmark

Attentional capture by schematic emotional faces was investigated in two experiments using the
flanker task devised by Eriksen and Eriksen (1974). In Experiment 1, participants were presented
with a central target (a schematic face that was either positive or negative) flanked by two identical
distractors, one on either side (schematic faces that were positive, negative, or neutral). The objective
was to identify the central target as quickly as possible. The impact of the flankers depended on their
emotional expression. Consistent with a threat advantage hypothesis (negative faces are processed
faster and attract more processing resources), responses to positive faces were slower when these were
flanked by (response incompatible) negative faces as compared with positive or neutral faces, whereas
responses to negative faces were unaffected by the identity of the flankers. Experiment 2 was a
standard flanker task with letter stimuli except that the task-neutral flankers were schematic faces
that were either positive, negative, or emotionally neutral. In this case, in which faces and emotional
expressions were to be ignored, performance seemed entirely unaffected by the faces. This result
suggests that attentional capture by emotional faces is contingent on attentional control settings.
Keywords: Visual attention; Attentional capture; Emotional expressions; Schematic faces; Flanker
task.

Although visual attention and emotion may

appear to be radically different processes, they
are closely related in an adaptive, functional sense.
The visual world presents us with a vast amount of
information at any given moment. Our visual
processing resources, however, are limited in
terms of processing and storage capacities, and
therefore must be directed to the most significant
parts of the visual field (see, e.g., Bundesen &
Habekost, 2008). Crucially, an important aspect
of significance is emotional relevance: that is,
relevance to our ongoing goals and concerns.

In the laboratory, the link between visual

attention and emotion has been investigated by a
number of experimental paradigms. A repeated
finding is that threatening stimuli (e.g., angry
faces) are capable of capturing attention in some
way. For example, a negative target face is
detected more quickly in a crowd of positive faces
than vice versa in a visual search task (e.g., Hansen
& Hansen, 1988), and a negative flanking face
causes more distraction than a positive flanking
face in a flanker task (Fenske & Eastwood, 2003;
Horstmann, Borgstedt, & Heumann, 2006). Such

Correspondence should be addressed to: Daniel Barratt, Department of International Culture and Communication Studies,
Copenhagen Business School, Dalgas Have 15, DK-2000 Frederiksberg, Denmark. E-mail: db.ikk@cbs.dk
# 2012 Psychology Press, an imprint of the Taylor & Francis Group, an Informa business
http://www.psypress.com/cogemotion

1223

http://dx.doi.org/10.1080/02699931.2011.645279

BARRATT AND BUNDESEN

findings are often supported by ecological and

evolutionary arguments: in the interests of selfpreservation, for example, threatening stimuli
need to be both attended to and responded to as
quickly as possible.
In the literature on visual attention, two
versions of the attentional capture hypothesis
can be distinguished (see Yantis, 1993, for a
discussion). The stimulus-driven capture hypothesis
states that attentional capture is driven by the
properties of the stimulus; that is, independently
of the current goals and concerns of the observer.
In this case, attentional control is said to be
bottom-up and exogenous. For example, experiments by Jonides and Yantis (1988) and
Theeuwes (1994) suggested that attention may
be automatically captured by either an abrupt
onset (the sudden appearance of a new stimulus)
or a featural singleton (a stimulus that differs
from adjacent objects in a single perceptual
dimension such as colour, brightness, or orientation). In contrast, the contingent capture hypothesis
states that attentional capture by a given stimulus
is contingent on the observers attentional control
settings or attentional set; that is, settings
determined by the requirements of the current
task. In this case, attentional control is said to be
top-down and endogenous. The strongest
version of this hypothesis states that this contingency applies to all visual features, including
abrupt onsets (see Folk, Remington, & Johnston,
1992). When considering emotional stimuli such
as faces, we are operating at the level of objects
rather than the level of features: for example, even
highly schematic facial expressions differ from
each other in several featural dimensions. Nevertheless, the basic question of whether attentional
capture by emotional objects is stimulus-driven or
contingent on attentional control settings still
applies: that is, is the observer looking out for
the angry face or not?
The current study addressed this question by
running two versions of the Eriksen flanker task
(Eriksen & Eriksen, 1974) with schematic faces as
emotional stimuli. In the first version, the objective was to categorise a centrally presented target
face as either positive (happy) or negative (sad/

1224

COGNITION AND EMOTION, 2012, 26 (7)

angry) and to ignore peripherally presented face

distractors; thus, the emotional dimension was
explicitly relevant to the task. In the second
version, the objective was to categorise a target
letter and to ignore both letter and face distractors; thus, the emotional dimension was explicitly
irrelevant to the task. For both versions of the
task, the influence of the participants attentional
control settings was interpreted at a qualitative
level in terms of the theory of visual attention
(TVA; Bundesen & Habekost, 2008)*a formal
computational theory originally proposed by
Bundesen (1990) and interpreted neurophysiologically by Bundesen, Habekost, and Kyllingsbk
(2005; also see Barratt, 2009). In short, TVA
proposes that visual attention comprises two
mechanisms of selection*filtering (selection of
objects)
and
pigeonholing
(selection
of
categories)*and involves two types of attentional
control parameters*pertinence values and bias
values. Pertinence values control the types of
objects that are attended, while bias values control
the types of categorisations that are made with
respect to those attended objects. Those objects/
categories which have the greatest pertinence and
bias are most likely to win the race for entry into
visual short-term memory (access consciousness).
The nature of attentional control settings is
much debated. Many investigators have emphasised the role of search templates (see, e.g.,
Duncan & Humphreys, 1989; Soto, Heinke,
Humphreys, & Blanco, 2005; Soto & Humphreys, 2007). In the theoretical account by
Desimone and Duncan (1995), Desimone
(1999), and Duncan (1996), visual search is
initiated by uploading a search template, which
is a representation of the target, into visual shortterm memory (VSTM; access consciousness)
which in turn yields a competition bias that
facilitates subsequent processing for sensory input
that matches the template. The account afforded
by TVA is different. In TVA, visual selection of
objects is determined by strengths of sensory
evidence (h values), perceptual biases (b values),
and pertinence (p) values. Other factors, such as
search templates, may influence the selection
process, but they do so via changes in h, b,

ATTENTIONAL CAPTURE BY EMOTIONAL FACES

and p values. The h values are computed by

comparing stimuli against representations in longor short-term memory. For example, the strength
of the sensory evidence that object x belongs to
category i, h(x, i), is computed by comparing
object x against a memory representation of
members of category i (the template associated
with category i). A representation in VSTM can
be used as a template for category i, that is, used
for computing h(x, i) by being compared against
x, but an observer does not necessarily search the
visual input for any item represented in VSTM
(cf. Bundesen, Habekost, & Kyllingsbk, 2011).

Previous studies of attentional capture by

emotional faces
A variety of experimental paradigms has been
used in studies on the relationship between visual
attention and emotion, including visual search and
the flanker task.1 The scope of this article is
limited to processing of faces with emotional
expressions. Below we present a brief review of
some of the major findings (see Frischen, Eastwood, & Smilek, 2008, for a more comprehensive
review of studies on visual search for faces with
emotional expressions; also see the review by
Yiend, 2010, of studies on the impact of emotion
on attention, and the recent study by Vogt, Lozo,
Koster, & De Houwer, 2010, on the role of goal
relevance in attentional orienting).
Visual search. The visual search task sheds light
on two key issues: the notion of attentional
(threat) capture, and the pros and cons of using
schematic faces as emotional stimuli. One of the
first studies was conducted by Hansen and
Hansen (1988) who presented their participants
with crowds of photographic faces (Ekman &
Friesen, 1976). The main finding was that
participants were significantly faster and more
accurate in detecting threatening (sad/angry) faces

in crowds of friendly (happy) faces than vice versa.

Subsequent researchers, however, have argued
that this threat advantage may have been due to
low-level perceptual confounds (see Purcell, Stewart, & Skov, 1996). In light of this, subsequent
versions of the visual search task have mainly used
schematic faces as emotional stimuli, the rationale
being that different emotional expressions can be
created from the same basic lines and shapes.
In most versions of the visual search task, the
participant is required to establish whether one of
the faces in the display is different from the rest
(target-present trials) or whether all of the faces
are the same (target-absent trials). In other
versions, the participant is always presented with
target-present displays and the task is to identify
the location of the target. Although the studies of
visual search with schematic faces have produced
variable results, some basic patterns, and corresponding hypotheses, can be distinguished.
One main hypothesis can be described as the
hman, Lundqvist,
threat advantage hypothesis (cf. O
& Esteves, 2001). By this hypothesis, threatening
faces are processed faster than non-threatening
faces. The hypothesis can explain the finding that
participants are both faster and more accurate at
detecting threatening (sad/angry) target faces than
non-threatening (happy) target faces in crowds of
distractor faces (for a representative sample of
studies, see Eastwood, Smilek, & Merikle, 2001;
hman et al.,
Fox et al., 2000; Nothdurft, 1993; O
2001; White, 1995; and but also see Horstmann,
2007, 2009; Horstmann, Becker, Bergmann, &
Burghaus, 2010).
hman (1993) proposed that feature detectors
O
perform preliminary stimulus analysis, before
passing the information to a significance system
and an arousal system. In the case of threatening
stimuli, the significance system can be bypassed
and the arousal system activated directly (cf.
LeDoux, 1995, 1998, 2000; Morris, DeGelder,
hman, &
Weiskrantz, & Dolan, 2001; Morris, O

1
Other experimental paradigms, not discussed here, include RSVP and attentional blink paradigms (e.g., Anderson & Phelps,
2001), the emotional Stroop task (Williams, Mathews, & MacLeod, 1996; also see MacLeod, 1991), and the spatial cuing (dotprobe) task (Bradley, Mogg, Falla, & Hamilton, 1998; MacLeod, Mathews, & Tata, 1986; also see Cooper & Langton, 2006; Fox,
Russo, Bowles, & Dutton, 2001; Fox, Russo, & Dutton, 2002).

COGNITION AND EMOTION, 2012, 26 (7)

1225

BARRATT AND BUNDESEN

hman, 2005; O
hman &
Dolan, 1998, 1999; O
Soares, 1994; Whalen et al., 2004).
Several recent studies have focused on the
distinction between automatic, stimulus-driven
capture by emotional faces and capture which is
contingent on attentional control settings. The
findings have varied to a large extent. Eastwood
et al. (2008) found evidence that interference from
emotional faces is not automatic but can be
eliminated and may depend on both attentional
control settings (e.g., for global vs. local processing) and stimulus attributes determining perceptual grouping. Blagrove and Watson (2010)
showed that previewing of schematic-face distractors facilitated visual search, but emotional
faces were difficult to ignore, and a full preview
benefit was not obtained with face stimuli. In an
MEG experiment, Fenker et al. (2010) found the
same N2pc component in response to irrelevant
fearful faces independent of the N2pc component
reflecting target selection and concluded that the
processing of the faces was automatic. Hodsoll,
Viding, and Lavie (2011) also claimed to show
automatic attentional capture by irrelevant emotional distractor faces in a search task in which the
emotion was entirely irrelevant. However, rather
than reflecting stimulus-driven attentional capture, their results may reflect a disengagement
effect (slower disengagement from emotional
faces) in a (possibly blind) serial search process
(cf. Horstmann & Becker, 2008).
Flanker task. A second experimental paradigm of
interest is the flanker task, originally developed by
Eriksen and Eriksen (1974). The objective of the
standard version of the flanker task is to categorise
a centrally presented target letter as quickly as
possible, while ignoring any distractor letters
presented on the left and right flanks. In the
simplest version of the task (see, e.g., Harms &
Bundesen, 1983), the central letter can be one of
two types (e.g., an H or an S). In the compatible
flanker condition, the flankers are the same as
the central target and thus compatible with the
required response (i.e., HHH, SSS). In the
incompatible flanker condition, the flankers are
different from the central target and associated

1226

COGNITION AND EMOTION, 2012, 26 (7)

with the opposite response (i.e., SHS, HSH).

In the neutral flanker condition, the flankers are
different from the central target and not associated with any response (e.g., OHO, OSO). The
general finding of the flanker task is described as
the flanker compatibility effect: Participants are
faster to respond to the central target when the
flankers are compatible as compared with incompatible. Normally, reaction times (RTs) to displays with compatible flankers are more or less the
same as RTs to displays without any flankers, RTs
to displays with neutral flankers are somewhat
slower than RTs to displays with compatible
flankers, and RTs to displays with incompatible
flankers are considerably slower than RTs to
displays with neutral flankers (see, e.g., Eriksen
& Eriksen, 1974; Harms & Bundesen, 1983).
The objective of the emotional version of the
flanker task is to categorise a centrally presented
target face as either positive (happy) or negative
(sad/angry), while ignoring any distractor faces
presented on the left and right flanks. The first
study (Fenske & Eastwood, 2003, Experiment
1A) used very simple schematic faces, while the
second study (Horstmann et al., 2006, Experiment 1) used the schematic faces designed by
hman et al. (see Figure 1). The compatible and
O
incompatible flanker conditions followed the same
logic as the letter version. In the neutral condition, the flanker faces had neutral expressions.
The main finding from the two studies was that
the response to the positive target was significantly affected by the type of flanker presented to
the left and right, whereas the response to the
negative target was similar across flanker conditions (see the upper histograms in Figure 1). The
most critical comparison was between the compatible and incompatible flanker conditions (see the
lower histograms in Figure 1): When the central
target was a positive face, the RTs for the
incompatible flanker condition (with negative
flankers) were significantly slower than the RTs
for the compatible flanker condition. When the
target was a negative face, on the other hand, the
RTs for the two conditions were more or less the
same. It is not clear why the effect of neutral
flankers differed between the two studies; in the

ATTENTIONAL CAPTURE BY EMOTIONAL FACES

Figure 1. Flanker task with schematic faces: Previous studies.

study of Fenske and Eastwood (2003), RTs to

displays with neutral flankers were essentially the
same as RTs to displays with incompatible
flankers, whether the central target was a positive
or a negative face (see, again, the upper histograms in Figure 1).
The threat advantage hypothesis also extends to
the flanker task: Negative faces attract attention to
a greater degree than positive faces. Thus, when
the target face is positive and the flanker faces are
negative, the participants attention is drawn
towards the flankers and away from the target,
thus causing an increase in the RT. When the
target face is negative and the flanker faces are

positive, on the other hand, the participants

response to the target is not significantly affected
by the flankers.
In the two experiments to be reported, we ran
two versions of the Eriksen flanker task with
schematic faces as emotional stimuli. Experiment
1 was largely a replication of the study by Fenske
and Eastwood (2003, Experiment 1A), and the
results supported the threat advantage hypothesis.
Experiment 2 was a standard flanker task with
letter stimuli except that the task-neutral flankers
were schematic faces that were either positive,
negative, or emotionally neutral. In this case, in
which faces and emotional expressions were
COGNITION AND EMOTION, 2012, 26 (7)

1227

BARRATT AND BUNDESEN

irrelevant to the task, performance seemed entirely

unaffected by the schematic faces. Thus, our
results suggest that attentional capture by schematic emotional faces is contingent on attentional
control settings.

EXPERIMENT 1
The effect of the flankers in the flanker task
depends on physical parameters of the stimulus
display such as the physical size of the stimuli and
the distance between the central target and the
flankers (see, e.g., Eriksen & Eriksen, 1974).
Experiment 1 was designed as a replication of
the study by Fenske and Eastwood (2003) with
one major change: In order to maximise the
potential effect of the flanking faces, the size of
the faces was increased, and the distance between
the central face and the flanking faces was
decreased.

Method
Participants. The participants were students from
the University of Copenhagen or high-school
students on a visit to the university. There was a
total of 40 participants (26 female, 14 male; age
range 18 to 37 years; mean age 21.1 years). The
university students were paid 120 Danish kroner
for their participation, while the high-school
students participated as part of an open day run
by the psychology department. Each participants
vision was either normal or corrected to normal.
Apparatus. The experiment was run on a number
of IBM-compatible desktop computers. The presentation of stimuli and the recording of responses
were controlled by E-Prime software (Psychology
Software Testing, Pittsburgh, PA). The stimuli
were displayed on a CRT monitor. The viewable
area was approximately 19 inches (48.3 cm)
measured diagonally and the resolution was set
at 800 600 pixels. The viewing distance was
approximately 60 cm. Responses were entered on
a standard keyboard.
Stimuli. The stimuli were schematic faces with
positive (happy), negative (sad/angry), and neutral

1228

COGNITION AND EMOTION, 2012, 26 (7)

expressions. The faces were taken from the

Wingdings collection of iconic symbols available
in many word-processing and graphics programs,
and were drawn in white against a black background. The faces had a diameter of 150 pixels
(:7 cm) and subtended a visual angle of around
6.78. The distance between the centre of the target
face and the centre of the flanking face was
175 pixels (:8.2 cm), giving an eccentricity of
around 7.88.
Procedure. The objective of the task was to
categorise a target face presented in the centre
of the display as either positive (happy) or
negative (sad/angry), while ignoring any
distractor faces presented on the left and right
flanks. The participants were instructed to respond as quickly as possible. The participants were
also told that there was likely to be a trade-off
between speed and accuracy and so they should
not worry if they made some mistakes: for
example, a 95% accuracy level (5% error rate)
was acceptable.
The experiment consisted of eight basic conditions: 2 Target Faces (positive and negative) 4
Flanker Conditions (no flanker, compatible, neutral, and incompatible; see Figure 2). In a single
block, each of the eight basic conditions was
repeated three times, giving a total of 24 trials.
The trials were randomly selected by the presentation software. The practice session consisted
of 1 block24 trials. The experimental session
consisted of 12 blocks 24 trials, giving a total of
288 trials. For each trial, a fixation cross was
presented for 1 s, followed without delay by the
presentation of the stimulus display. The presentation of the stimulus display was terminated by
the participants response. A feedback display*
providing information on accuracy and RT*was
presented for 1.5 s in order to orient and motivate
the participant.
The response keys were c and m on the lowest
row of letters. The participants were instructed to
rest their left index finger on the c key and their
right index finger on the m key. The keys were
labelled either  for a positive face or
for a
negative face. In order to account for the effects of

ATTENTIONAL CAPTURE BY EMOTIONAL FACES

Figure 2. Conditions in Experiment 1.

handedness, the participants were divided into

two groups of 20: the first group identified
positive targets with their left index finger and
negative targets with their right index finger,
whereas the second group identified negative
targets with their left index finger and positive
targets with their right index finger. Each group
contained one left-handed participant.

Results
Both RT outliers (0.5% of the total trials) and
incorrect responses (6.5%) were excluded from the
analysis. In order to determine outliers, a simple
cut-off method was employed: that is, RTs either
shorter than 200 ms or longer than 1,000 ms were
removed. The mean RTs for the correct responses
for each of the eight conditions are presented in
Figure 3.
The first statistical analysis was done using a 2
(Target Emotion: positive vs. negative) 2 (Flanker Condition: compatible vs. incompatible) repeated-measures analysis of variance (ANOVA).
In accordance with the standard Eriksen flanker

task, the main effect of Flanker Condition on

mean RTs was significant, F(1, 39) 6.15,
MSE271.96, pB.05. The main effect of Target
Emotion on mean RTs was not significant; that is,
there was no significant difference between the
mean RTs for positive and negative faces when
the data were collapsed across the two flanker
conditions, FB1. Importantly, however, there
was a significant interaction between Target
Emotion and Flanker Condition; that is, the
target emotion had a significant effect on the
relationship between the two flanker conditions,
F(1, 39) 13.16, MSE 182.94, pB.001. The
nature of this interaction was investigated by
running two paired-sample t-tests. When the
central target was a positive face, the mean RT
for the incompatible flanker condition was significantly slower than the mean RT for the
compatible condition (439 ms vs. 425 ms),
t(39)3.62, pB.001. Conversely, when the central target was a negative face, the mean RTs for
the incompatible and the compatible conditions
were essentially the same (432 ms vs. 433 ms),
t(39)0.48, p.64.2

We also conducted an overall analysis using a 2 (Target Emotion: positive vs. negative) 4 (Flanker Condition: no-flanker,
compatible, neutral, incompatible) repeated-measures ANOVA. This analysis revealed a similar pattern of results: The main effect
of Flanker Condition on mean RTs was significant, F(3, 117)  7.42, MSE 263.61, p B.001; the main effect of Target Emotion
was not significant, F B1; and the interaction between Target Emotion and Flanker Condition was significant, F(2.64,
102.81)  5.15, MSE 239.18, p B.01. For all of the multi-factor ANOVAs reported in this paper, the Greenhouse
Geisser
correction was used to adjust the degrees of freedom in the cases where Mauchlys test indicated that the assumption of sphericity
had been violated.
COGNITION AND EMOTION, 2012, 26 (7)

1229

BARRATT AND BUNDESEN

Figure 3. Mean reaction times (RTs) for Experiment 1. The mean error rates for each condition (in percent) are presented in parentheses at
the top of the chart.

An analysis of the corresponding error rates

(after the RT outliers had been removed) revealed
a similar pattern of results. The main effect of
Flanker Condition was significant, F(1,
39) 11.83, MSE0.002, pB.01, while the
main effect of Target Emotion was not significant, F B1. The interaction between the two
factors was significant, F(1, 39) 5.48, MSE
0.002, pB.05. For positive targets, more errors
were made for the incompatible than the compatible condition (8.9% vs. 4.9%), t(39) 4.35,
pB.001. For negative targets, there was no
significant difference between the two conditions
(7.2% vs. 6.5%), t(39)0.70, p.49. This comparison suggests that there was not a speed

accuracy trade-off; in other words, faster responses
were not sacrificed for greater accuracy.
The aim of the second analysis was to ascertain
whether there was a flanker presence effect, and
whether there was any difference between the
processing speed for the positive and negative
targets. This analysis was done by comparing the
no-flanker and neutral conditions. A 2 (Target
Emotion: positive vs. negative) 2 (Flanker Condition: no-flanker vs. neutral) repeated-measures
ANOVA revealed a significant flanker presence

1230

COGNITION AND EMOTION, 2012, 26 (7)

effect, such that RTs were faster in the no-flanker

condition, F(1, 39) 8.70, MSE267.58,
pB.01. The main effect of Target Emotion was
not significant, F(1, 39) 2.20, MSE572.08,
p.15, nor was the interaction between Target
Emotion and Flanker Condition, F(1, 39) 1.72,
MSE262.29, p.20. An analysis of the corresponding error rates revealed no significant main
effects or interactions, Fs B1.

Discussion
The results of Experiment 1 can be broken down
into two key findings, one concerning the flanker
dimension and one concerning the emotional
dimension. First, the main effect of flanker
condition on mean RTs was significant, suggesting that the task as a whole was operating in the
way predicted by the standard account of performance in the flanker task paradigm. Second, and
even more importantly, the interaction between
flanker condition and target emotion was also
significant, suggesting that the emotional identity
of both the target and the flanking faces had an
impact on the participants performance. The
analysis of the compatible and incompatible

ATTENTIONAL CAPTURE BY EMOTIONAL FACES

conditions revealed that the response to

the positive target was significantly slower when
the flankers were negative, whereas the response
to the negative target was not affected by the
identity of the flankers. Following the threat
advantage hypothesis, this finding can be explained by assuming that negative faces attract
more processing resources and are processed at a
faster rate than positive faces. In the key condition, therefore, the negative flankers drew the
participants attention away from the positive
target, thereby slowing their response.
The analysis of the no-flanker and neutral
conditions revealed two key points. First, both
positive and negative targets were processed faster
when no flankers were presented to the left and
right. This result can be explained by assuming
that all of the participants attentional processing
resources could be devoted to the central target
when this was presented alone. Second, the main
effect of target emotion was not significant.
Indeed, in the no-flanker condition, RTs to
positive and negative faces differed by only 3 ms
while the corresponding error rates differed by
only 0.6%.

TVA-based interpretation. In terms of the theory

of visual attention (TVA; Bundesen & Habekost,
2008), a qualitative threat advantage interpretation of Experiment 1 can be formulated as follows.
In order to perform the flanker task, the participant ascribes high pertinence (p value) to the
target location so that objects that appear at the
target location receive high attentional weights
(cf. the weight equation of TVA). Certain features
other than location are also ascribed pertinence. In
particular, when emotional expressions of faces are
task relevant, as was the case in Experiment 1,
physical features of emotional faces are given high
pertinence values, so that emotional faces receive
high attentional weights. For ecological reasons,
however, pertinence values of faces with negative
emotional expressions are higher than pertinence
values of faces with positive emotional expressions, whence attentional weights of faces with
negative expressions are higher than attentional

weights of faces with positive expressions. Accordingly, by the rate equation of TVA, the rate
(v) at which negative faces are processed is faster
than the rate at which positive faces are processed.
For this reason, positive faces flanking a negative
target face have little effect on responses to the
negative target face, whereas negative faces flanking a positive target face interfere with the
response to the target.
In addition to being favoured by higher
pertinence values, processing of negative faces
might conceivably also be favoured by higher
perceptual bias (b) values. However, if this had
been the case in Experiment 1, responses to
negative faces in the no-flanker condition should
have been faster and more accurate than responses
to positive faces in the same condition, which was
not the case.

EXPERIMENT 2
The results of Experiment 1 provide further
evidence for the hypothesis that negative (sad/
angry) faces capture attention more than positive
(happy) faces, or in TVA terms, that negative
faces have higher pertinence values and
attentional weights than positive faces. The
experimental design, however, did not allow us
to make a distinction between stimulus-driven
and contingent capture (cf. Folk et al., 1992).
Given that participants had to identify the faces
on the basis of emotional expression (positive or
negative), the threatening component of the
negative faces was of relevance to the task in
hand. In order to test the hypothesis that task
relevance is of importance, a mixed version of the
flanker task was run with letter targets and letter/
face distractors, thus combining the standard
Eriksen flanker task using letter stimuli with the
emotional flanker task using face stimuli.
In the revised version of the task, the objective
was to categorise a letter presented in the centre of
the display as quickly as possible and to ignore any
distractors presented on the left and right flanks.
The no-flanker, compatible, and incompatible
conditions were the same as for the standard
COGNITION AND EMOTION, 2012, 26 (7)

1231

BARRATT AND BUNDESEN

Eriksen flanker task described above. In the taskneutral flanker condition, however, the flankers
were schematic faces (with positive, negative, or
neutral expressions) as opposed to task-neutral
letters. By definition, then, the schematic faces
were irrelevant to the task set for the participant.
In a pilot study, the letters H and S were used as
targets. A preliminary analysis of the results,
however, suggested that there might be some
confusion between the horizontal line of the H
and the horizontal mouth of the neutral face. In
the main study, therefore, the letter H was
replaced with the letter N, or, to describe the
operation in terms of basic stimulus features, the
horizontal line joining the two vertical bars was
replaced with a diagonal line.

Method
Participants. Thirty staff members and students
at the University of Copenhagen volunteered to
participate in the experiment (15 females, 15
males; age range 22 to 56 years; mean age 35.7
years). Each participant had either normal or
corrected-to-normal vision.
Stimuli. The face stimuli were the same as those
used in Experiment 1: namely, schematic faces
with positive (happy), negative (sad/angry), and
neutral expressions. The letter stimuli were drawn
in an Arial font and presented inside the circular
outline of the schematic face. This was to ensure
that participants could not filter out the face
distractors on the basis of a single and relatively
salient stimulus feature. Both types of stimuli were
drawn in white against a black background. Both

Figure 4. Conditions in Experiment 2.

1232

COGNITION AND EMOTION, 2012, 26 (7)

the diameter of the stimuli and the distance

between target and flankers were the same as in
Experiment 1.
Procedure. The objective was to categorise a
target letter presented in the centre of the display
as either an N or an S, and to ignore any
distractor stimuli presented on the left and right
flanks. The verbal and written instructions regarding speed
accuracy trade-off were the same as
in Experiment 1. The experiment comprised 12
basic conditions in total: 2 Target Letters (the
letters N and S)6 Flanker Conditions (the noflanker, compatible, and incompatible conditions
in conjunction with three versions of the neutral
condition; see Figure 4). In a single block, each of
the six basic flanker conditions was repeated three
times, while each of the six neutral flanker
conditions was shown only once, giving a total
of 24 trials. The trials were randomly selected by
the presentation software. The practice session
consisted of 1 block 24 trials. The experimental
session consisted of 12 blocks 24 trials, giving a
total of 288 trials. Other details of the procedure
(regarding fixation and feedback) were the same as
in Experiment 1.
The c and m keys on the lowest row of letters
were labelled either N or S. In order to
account for the effects of handedness, the participants were divided into two groups of 15: the
first group identified the letter N with their left
index finger and the letter S with their right index
finger, whereas the second group identified the
letter S with their left index finger and the letter

ATTENTIONAL CAPTURE BY EMOTIONAL FACES

Figure 5. Mean reaction times (RTs) for Experiment 2. For the sake of simplicity, the mean RTs for the two target letters have been pooled
together. The mean error rates for each condition (in percent) are presented in parentheses at the top of the chart.

N with their right index finger. Each group

contained one left-handed participant.

Results
Both RT outliers (0.8% of the total trials) and
incorrect responses (5.9%) were excluded from the
analysis. The mean RTs for the correct responses
for each of the six flanker conditions are presented
in Figure 5. For reasons outlined below, the mean
RTs for target letters N and S are considered
together.
The aim of the first statistical analysis was to
test the relationship between the letter-only
conditions, in order to establish that the task
was working in the expected way. This analysis
was done using a 2 (Target Letter: N vs. S)3
(Flanker Condition: no-flanker vs. compatible vs.
incompatible) repeated-measures ANOVA. In
accordance with results from previous studies
with the standard Eriksen flanker task, the main
effect of Flanker Condition on mean RTs was
significant, F(2.00, 57.95) 8.27, MSE219.42,
pB.001. In contrast, neither the main effect of
Target Letter nor the interaction between Target

Letter and Flanker Condition were significant,

Fs B 1. Thus, although letter identity was relevant to the task, there was no difference in RT
performance between the target letters. An analysis of the corresponding error rates (after the RT
outliers had been removed) revealed a similar
pattern of results, with only the main effect of
Flanker Condition reaching significance, F(1.87,
54.33) 6.62, MSE 0.002, pB.01.
The aim of the second statistical analysis was to
test the relationship between the three neutral
conditions with schematic faces as flankers. This
analysis was done using a 2 (Target Letter: N vs.
S)3 (Flanker Emotion: positive vs. neutral vs.
negative) repeated-measures ANOVA. The main
effect of Flanker Emotion was not significant,
F(1.99, 57.83) 2.03, MSE296.70, p .14,
with the mean RTs for displays with positive,
neutral, and negative faces being 387 ms, 388 ms,
and 382 ms, respectively. As in the letter-only
conditions, neither the main effect of Target
Letter (FB1) nor the interaction between Target
Letter and Flanker Emotion, F(1.70, 49.39)
1.07, MSE512.93, p.34, were significant. An
COGNITION AND EMOTION, 2012, 26 (7)

1233

BARRATT AND BUNDESEN

analysis of the corresponding error rates revealed a

similar pattern of results, Fs B1.
The objective of the third and final analysis was
to test the relationship between the letter-only
conditions and the neutral condition with schematic faces as flankers. Because the preceding
analyses showed no significant difference between
the mean RTs for the two target letters, the RTs
for the two target letters were pooled together.
Because positive and neutral faces are not central
to the threat advantage hypothesis, only the
neutral condition with negative faces was considered. A paired-sample t-test revealed that there
was no significant difference between the mean
RT for the neutral condition with negative faces
(382 ms) and the mean RT for the no-flanker
condition (383 ms), t(29) 0.42, p .68. The
difference between the corresponding error rates
(4.2% vs. 5.3%, respectively) was also not significant, t(29)1.17, p .25. In short, negative
faces had no more impact on the participants
performance than empty circles.

Discussion
As in the case of Experiment 1, the results of
Experiment 2 can be broken down into two key
findings, one concerning the flanker dimension
and one concerning the emotional dimension.
First, the analysis of the letter-only conditions
confirmed that the pattern of mean RTs conformed to expectations from previous studies
using the standard flanker task of Eriksen and
Eriksen (1974). In particular, RTs to displays with
incompatible letter flankers were slower than RTs
to displays with compatible flankers or without
any flankers. Second, and even more importantly,
the analysis of the neutral condition with flanking
faces suggested that attentional capture by negative (sad/angry) faces is not automatic, but contingent on attentional control settings (cf. Folk
et al., 1992). In contrast to the faces-only version
of the flanker task, the negative faces had no
significant effect on performance, suggesting that
task relevance is crucial. Most tellingly, there was
no significant difference between the mean RT for
the neutral condition with negative faces and

1234

COGNITION AND EMOTION, 2012, 26 (7)

the mean RT for the no-flanker condition with

empty circles.

GENERAL DISCUSSION
A repeated finding in the literature on visual
attention and emotion is that threatening stimuli
are capable of capturing attention in some way.
The results of Experiment 1*the flanker task
with schematic faces*support the hypothesis that
negative (sad/angry) faces attract more processing
resources and are processed faster than positive
(happy) faces, or, in TVA terms, that negative
faces have higher pertinence values and attentional weights than positive faces. Because the
emotional identity of the faces was relevant to the
task, however, the experimental design did not
make a distinction between stimulus-driven and
top-down controlled capture. In contrast, the
results of Experiment 2*the flanker task with
letter targets and letter/face distractors*suggest
that attentional capture by negative faces is not
automatic, but dependent on attentional control
settings. When the task was to detect a target
letter rather than a target face, negative flanking
faces had no significant effect on performance. In
TVA terms, the attentional weights of positive
and negative faces appeared to be as low as the
attentional weights of empty circles.
In summary, Experiments 1 and 2 have
provided evidence for two previously suggested
principles of attention for emotional facial expressions, formulated in the threat advantage
hypothesis and the contingent capture hypothesis,
respectively. By the threat advantage hypothesis,
threatening stimuli like negative faces attract
more processing resources and are processed faster
than otherwise similar non-threatening stimuli.
By the contingent capture hypothesis, attentional
capture by a given stimulus is contingent on the
observers attentional control settings. This hypothesis seems true for emotional faces, and
according to TVA, the hypothesis is true for all
types of stimuli.

ATTENTIONAL CAPTURE BY EMOTIONAL FACES

Open questions
The assumption that attentional capture by emotional faces is contingent on attentional control
settings was supported by Experiments 1 and 2:
Emotional faces captured attention when faces
and emotional expressions were relevant to the
task (Experiment 1) but not when faces and
emotional expressions were irrelevant to the task
(Experiment 2). Precisely which aspects of task
relevance determined the contrast in performance
between Experiments 1 and 2 is less clear.
Presumably, the fact that emotional faces appeared as targets in Experiment 1, whereas
participants in Experiment 2 never responded to
emotional faces, was a major determinant of the
contrast in performance between the two experiments, but other factors may also have been
important. For example, the perceptual similarity
of the affective flankers to the targets (i.e., the
resemblance of the affective flankers to the task
relevant stimuli) was arguably smaller in Experiment 2 as compared to Experiment 1, which
conceivably may have diminished the effect of the
affective flankers in Experiment 2.
A limitation in the current study is that
individual differences in trait anxiety were not
taken into account. A negative schematic face may
be regarded as a mild threat. Consistent with
Mogg and Bradleys (1998) cognitive-motivational analysis of anxiety, suppose low anxious
people show attentional avoidance of mild threats,
whereas high anxious people show attentional
attraction towards threat (cf. Bar-Haim, Lamy,
Pergamin, Bakermans-Kranenburg, & van Yzendoorn, 2007). If so, the two effects might
conceivably cancel out in mean RTs and produce
a null effect of negative flankers as observed in
Experiment 2. This possibility may be tested in
future studies by including, for example, a questionnaire measuring individual differences in
anxiety. Measuring individual differences in anxiety would also make it possible to match participants in different experiments closely with respect
to (state, trait, or social) anxiety. As matters stand,
the difference in results between Experiments 1
and 2 might conceivably be due to differences

between the participants. In Experiment 1, participants were high school and university students
with a mean age of 21 years, and 65% of the
participants were female, while in the second
study participants were university students and
staff with a mean age of 36 years, and only 50% of
the participants were female. It is possible that
participants in Experiment 1 had higher levels of
anxiety than participants in Experiment 2, and
this may have contributed to the results.
Manuscript received 29 May
Revised manuscript received 30 October
Manuscript accepted 14 November
First published online 13 March

2011
2011
2011
2012

REFERENCES
Anderson, A. K., & Phelps, E. A. (2001). Lesions of
the human amygdala impair enhanced perception of
emotionally salient events. Nature, 411, 305
309.
Bar-Haim, Y., Lamy, D., Pergamin, L., BakermansKranenburg, M. J., & van Yzendoorn, M. H.
(2007). Threat-related attentional bias in anxious
and non-anxious individuals: A meta-analytic study.
Psychological Bulletin, 133, 1
24.
Barratt, D. (2009). Investigating the link between visual
attention and emotion: A TVA-based computational approach. Journal of Vision, 9(8), abstract 199.
Blagrove, E., & Watson, D. G. (2010). Visual marking
and facial affect: Can an emotional face be ignored?
Emotion, 10, 147
168.
Bradley, B., Mogg, K., Falla, S. J., & Hamilton, L. R.
(1998). Attentional bias for threatening facial
expressions in anxiety: Manipulation of stimulus
duration. Cognition and Emotion, 12, 737
753.
Bundesen, C. (1990). A theory of visual attention.
Psychological Review, 97, 523
547.
Bundesen, C., & Habekost, T. (2008). Principles of
visual attention: Linking mind and brain (Oxford
Psychology Series). Oxford, UK: Oxford University
Press.
Bundesen, C., Habekost, T., & Kyllingsbk, S. (2005).
A neural theory of visual attention: Bridging cognition and neurophysiology. Psychological Review, 112,
291
328.
Bundesen, C., Habekost, T., & Kyllingsbk, S. (2011).
A neural theory of visual attention and short-term
memory. Neuropsychologia, 49, 1446
1457.
COGNITION AND EMOTION, 2012, 26 (7)

1235

BARRATT AND BUNDESEN

Cooper, R. M., & Langton, S. R. H. (2006). Attentional bias to angry faces using the dot-probe task? It
depends when you look for it. Behavioural Research
and Therapy, 44, 1321
1329.
Desimone, R. (1999). Visual attention mediated by
biased competition in extrastriate visual cortex. In
G. W. Humphreys, J. Duncan, & A. M. Treisman
(Eds.), Attention, space, and action (pp. 13
30).
Oxford, UK: Oxford University Press.
Desimone, R., & Duncan, J. (1995). Neural mechanisms of selective visual attention. Annual Review of
Neuroscience, 18, 193
222.
Duncan, J. (1996). Cooperating brain systems in
selective perception and action. In T. Inui & J. L.
McClelland (Eds.), Attention and performance XVI:
Information integration in perception and communication (pp. 549
578). Cambridge, MA: MIT Press.
Duncan, J., & Humphreys, G. W. (1989). Visual search
and stimulus similarity. Psychological Review, 96,
433
458.
Eastwood, J. D., Frischen, A., Reynolds, M., Gerritsen,
C., Dubins, M., & Smilek, D. (2008). Do emotionally expressive faces automatically capture attention? Evidence from global
local interference.
Visual Cognition, 16, 248
261.
Eastwood, J. D., Smilek, D., & Merikle, P. M. (2001).
Differential attentional guidance by unattended
faces expressing positive and negative emotion.
Perception & Psychophysics, 63, 1004
1013.
Ekman, P., & Friesen, W. V. (1976). Pictures of facial
affect. Palo Alto, CA: Consulting Psychologists
Press.
Eriksen, B. A., & Eriksen, C. W. (1974). Effects of
noise letters upon the identification of a target letter
in a nonsearch task. Perception and Psychophysics, 16,
143
149.
Fenker, D. B., Heipertz, D., Boehler, C. N., Schoenfeld, M. A., Noesselt, T., Heinze, H.-J., et al.
(2010). Mandatory processing of irrelevant fearful
face features in visual search. Journal of Cognitive
Neuroscience, 22, 2926
2938.
Fenske, M. J., & Eastwood, J. D. (2003). Modulation
of focused attention by faces expressing emotion:
Evidence from flanker tasks. Emotion, 3, 327
343.
Folk, C. L., Remington, R. W., & Johnston, J. C.
(1992). Involuntary covert orienting is contingent on
attentional control settings. Journal of Experimental
Psychology: Human Perception and Performance, 18,
1030
1044.
Fox, E., Lester, V., Russo, R., Bowles, R. J., Pichler, A.,
& Dutton, K. (2000). Facial expressions of emotion:

1236

COGNITION AND EMOTION, 2012, 26 (7)

Are angry faces detected more efficiently? Cognition

and Emotion, 14(1), 61
92.
Fox, E., Russo, R., Bowles, R., & Dutton, K. (2001).
Do threatening stimuli draw or hold visual attention
in subclinical anxiety? Journal of Experimental
Psychology: General, 130, 681
700.
Fox, E., Russo, R., & Dutton, K. (2002). Attentional
bias for threat: Evidence for delayed disengagement
from emotional faces. Cognition and Emotion, 16,
355
379.
Frischen, A., Eastwood, J. D., & Smilek, D. (2008).
Visual search for faces with emotional expressions.
Psychological Bulletin, 134, 662
676.
Hansen, C. H., & Hansen, R. D. (1988). Finding the
face in the crowd: An anger superiority effect.
Journal of Personality and Social Psychology, 54,
917
924.
Harms, L., & Bundesen, C. (1983). Color segregation
and selective attention in a nonsearch task.
Perception & Psychophysics, 33, 11
19.
Hodsoll, S., Viding, E., & Lavie, N. (2011). Attentional capture by irrelevant emotional distractor
faces. Emotion, 11, 346
353.
Horstmann, G. (2007). Preattentive face processing:
What do visual search experiments with schematic
faces tell us? Visual Cognition, 15, 799
833.
Horstmann, G. (2009). Visual search for schematic
affective faces: Stability and variability of search
slopes with different instances. Cognition and
Emotion, 23, 355
379.
Horstmann, G., & Becker, S. I. (2008). Attentional
effects of negative faces: Top-down-contingent or
involuntary? Perception & Psychophysics, 70,
1416
1434.
Horstmann, G., Becker, S. I., Bergmann, S., &
Burghaus, L. (2010). A reversal of the search
asymmetry favouring negative schematic faces.
Visual Cognition, 7, 981
1016.
Horstmann, G., Borgstedt, K., & Heumann, M.
(2006). Flanker effects with faces may depend on
perceptual as well as emotional differences. Emotion,
6, 28
39.
Jonides, J., & Yantis, S. (1988). Uniqueness of abrupt
visual onset in capturing attention. Perception &
Psychophysics, 43, 346
354.
LeDoux, J. E. (1995). In search of an emotional system
in the brain: Leaping from fear to emotion and
consciousness. In M. S. Gazzaniga (Ed.), The
cognitive neurosciences (pp. 1049
1061). Cambridge,
MA: MIT Press.

ATTENTIONAL CAPTURE BY EMOTIONAL FACES

LeDoux, J. E. (1998). The emotional brain: The

mysterious underpinnings of emotional life. London,
UK: Weidenfeld & Nicolson.
LeDoux, J. E. (2000). Cognitive emotional interactions:
Listen to the brain. In R. D. Lane & L. Nadel
(Eds.), Cognitive neuroscience of emotion (pp. 129

155). New York, NY: Oxford University Press.
MacLeod, C. M. (1991). Half a century of research on
the Stroop effect*An integrative review. Psychological Bulletin, 109, 163
203.
MacLeod, C. M., Mathews, A., & Tata, P. (1986).
Attentional bias in emotional disorders. Journal of
Abnormal Psychology, 95, 15
20.
Mogg, K., & Bradley, B. P. (1998). A cognitivemotivational analysis of anxiety. Behaviour Research
and Therapy, 36, 809
848.
Morris, J. S., DeGelder, B., Weiskrantz, L., & Dolan,
R. J. (2001). Differential extrageniculostriate and
amygdala responses to presentation of emotional
faces in a cortically blind field. Brain, 124,
1241
1252.
hman, A., & Dolan, R. J. (1998).
Morris, J. S., O
Conscious and unconscious emotional learning in
the human amygdala. Nature, 393, 467
470.
hman, A., & Dolan, R. J. (1999). A
Morris, J. S., O
subcortical pathway to the right amygdala mediating
unseen fear. Proceedings of the National Academy of
Sciences USA, 96, 1680
1685.
Nothdurft, H. C. (1993). Faces and facial expressions
do not pop out. Perception, 22, 1287
1298.
hman, A. (1993). Fear and anxiety as emotional
O
phenomena: Clinical phenomenology, evolutionary
perspectives, and information processing. In
M. Lewis & J. M. Haviland-Jones (Eds.), Handbook
of emotions (pp. 511
536). New York, NY: Guilford
Press.
hman, A. (2005). The role of the amygdala in human
O
fear: Automatic detection of threat. Psychoneuroendocrinology, 30, 953
958.
hman, A., Lundqvist, D., & Esteves, F. (2001). The
O
face in the crowd revisited: A threat advantage with

schematic stimuli. Journal of Personality and Social

Psychology, 80, 381
396.
hman, A., & Soares, J. J. F. (1994). Unconscious
O
anxiety: Phobic responses to masked stimuli. Journal of Abnormal Psychology, 103, 231
240.
Purcell, D. G., Stewart, A. L., & Skov, R. B. (1996). It
takes a confounded face to pop out of a crowd.
Perception, 25, 1091
1108.
Soto, D., Heinke, D., Humphreys, G. W., & Blanco,
M. J. (2005). Early, involuntary top-down guidance
of attention from working memory. Journal of
Experimental Psychology: Human Perception and Performance, 31, 248
261.
Soto, D., & Humphreys, G. W. (2007). Automatic
guidance of visual attention from verbal working
memory. Journal of Experimental Psychology: Human
Perception and Performance, 33, 730
737.
Theeuwes, J. (1994). Stimulus-driven capture and
attentional set: Selective search for color and visual
abrupt onsets. Journal of Experimental Psychology:
Human Perception and Performance, 20, 799
806.
Vogt, J., Lozo, L., Koster, E. H. W., & De Houwer, J.
(2010). On the role of goal relevance in emotional
attention: Disgust evokes early attention to cleanliness. Cognition and Emotion, 25, 466
477.
Whalen, P. J., Kagan, J., Cook, R. G., Davis, F. C.,
Kim, H., Polis, S., et al. (2004). Human amygdala
responsivity to masked fearful eye whites. Science,
306, 2061.
White, M. (1995). Preattentive analysis of facial
expressions of emotion. Cognition and Emotion, 9,
439
460.
Williams, J. M. G., Mathews, A., & MacLeod, C.
(1996). The emotional Stroop task and psychopathology. Psychological Bulletin, 120, 3
24.
Yantis, S. (1993). Stimulus-driven attentional capture
and attentional control settings. Journal of Experimental Psychology: Human Perception and Performance, 19, 676
681.
Yiend, J. (2010). The effects of emotion on attention: A
review of attentional processing of emotional information. Cognition and Emotion, 24, 3
47.

COGNITION AND EMOTION, 2012, 26 (7)

1237

Copyright of Cognition & Emotion is the property of Psychology Press (UK) and its content may not be copied
or emailed to multiple sites or posted to a listserv without the copyright holder's express written permission.
However, users may print, download, or email articles for individual use.